It’s 02020-11-15. Four days ago my options for microcontroller output were limited to one four-digit LED display I’d painstakingly desoldered from a microwave, some other LEDs I’d painstakingly desoldered from other random discarded electronic equipment, and the microwave’s beeper. Now the covid quarantine is over and I’ve been able to visit my apartment (I’ve spent the last 7½ months at my girlfriend’s apartment) and now I have a number of additional options.
| display | w(mm) | h(mm) | type | shows | color | pins | conn |
|------------+-------+-------+------+---------+-------+------+-------------|
| DeV96 | 50 | 18 | LED | 4 0-9 | red? | 14 | SIP 2.54mm |
| microwave | 45 | 12 | LED | 4 0-9 | green | 14 | SIP 2.54mm |
| breadboard | 50 | 18 | LED | 4 0-9 | red | 36 | DIP 2.54mm |
| Kenko | 52 | 12 | LCD | 8 0-9 | grey | ≈32 | flex 1.26mm |
| Kadio | 60 | 18 | LCD | 12 a-z? | grey | ≈75 | flex 0.8mm |
| | | | | +12 0-9 | | | |
| Franklin | 53 | 18 | LCD | 25 a-z? | grey | ≈105 | flex 0.5mm |
Some of these are sort of package deals with a keyboard and maybe a battery and case, which tempts me to rebrain them (see Rebraining), while others are just displays.
The LED displays are suitable for PWM fading; not sure if this will work with the LCDs. Nobody ever does PWM fading of 7-segment LED displays, which is going to rock. Also I think LED displays would look much better shining through cloth, another thing nobody ever does. These LCDs are all reflective LCDs, so they would have the best readability in daylight.
Cellphone screens are pretty interesting, especially SPI Nokia screens, although I don’t happen to have any at the moment.
A guy at a hackerspace gifted me this display he’d made on perfboard, evidently on 01996-10-13. It has an 8-key keypad, apparently configured as a 2×4 matrix with diodes, and five BC548B transistors onboard; four of them are hooked up to turn on a particular digit of the 7-segment LED display. I traced the wiring out before, but I forget if you can fully control the display from offboard with the 14-pin header he’s provided, or if it’s partly under control of the keyboard.
The 7-segment displays are helpfully in a socket, so not only can you replace them if you burn them out, but you can easily transplant them into other things. Each has a decimal point.
This may be the easiest option to use, since it already has current-limiting resistors (probably sized for 5V) and common-electrode transistors with current-limiting resistors on it. The individual segments are driven from offboard, so you need to source or sink (source I think) enough current to light them up. It’s also one of the largest displays, which will make it easy to read. It’s probably not using high-brightness LEDs, which means it won’t be as bright as some other options, especially if run off 2V or 3V.
I suspect 12 pins of the 14 pins, plus power and ground, would be enough to drive the display, and it would probably work down to 2V.
The front-panel display from the microwave oven I ripped apart has four 7-segment digits without decimal points, a colon with separately illuminable dots, and ten miscellaneous microwave-oven-related ideographic indicators above and below. The digits are slightly smaller than the DeV96 digits, but the component as a whole is much smaller, and it lacks any onboard resistors or transistors. There are nine cathodes (?) for the individual segments of a digit, then five common anodes (?) to select the digit; one of the five “digits” is the colon and two of the miscellaneous indicators, while each of the other digits controls two more miscellaneous indicators. So you can control all the digits proper with 11 pins.
The digit segments light up faintly green on the milliamp or two from my multimeter’s diode-test mode, but are presumably intended for a much higher current. I could probably stress one of the colon dots until it blows to get an idea of what they could stand, but 20mA is probably safe, and close to the limit of what my microcontrollers can provide anyway.
It’s labeled GAL9801-OI on one side, but GAL-something was the microwave’s model number, and no datasheets seem to be forthcoming.
I was able to get the microwave’s somewhat yellowed front panel off in one piece; it has a flat-flex cable (dated “11/18/99”) which went into a 12-pin cable-pinch connector on the mainboard (where the display was soldered), and I was able to desolder that connector, which has the same breadboard-friendly 2.54mm pin spacing as the display. Its polarized front window of course can fit the display in it. It has 17 membrane-keyboard buttons labeled with microwave-related things and connected in some kind of matrix with resistive ink on flat-flex to the 12-pin connector. I was able to detect a press on one of them with the multimeter; it read as about 120 Ω.
I’m not sure, but I think these are “bright green” 3-volt LEDs rather than the older 1.7-volt-or-so older green type. So running the display at 20 mA per segment would probably cost 420 mW worst-case. But at an equivalent brightness it probably costs less power than the DeV96 display.
I could very reasonably tack the display in place inside the front panel with some silicone or something, then mount a control board and even power supply inside, repurposing the shitty membrane microwave keyboard for some nobler purpose, perhaps covering it with labels. The 7 segment-selector pins could presumably be shared with at least 6 and probably 7 of the pins for the keyboard, so you’d probably only need 16 pins between the keyboard and the display. The 18-GPIO ATTiny2313 has enough pins for that, but not to do much else, and no ADC, which would be useful for many of the things I want to show on the display. But it can blow or suck 40 mA per I/O pin, so it could drive the display at a dimly readable intensity without any external hardware at all, and probably very brightly if supplied with four per-digit high-side transistors.
If I’m wrong and they’re actually low-side transistors, it might be able to drive the transistors with its internal pullup resistors! 50kΩ worst-case at 5 volts gives us 100 μA, and with an unremarkable β of 300, that would give us... 30 mA. Which is worse than what it can do itself. sad trombone It might happen to be higher than that, or you might be able to use a darlington. But then you can also get transistors with integrated base resistors.
An STM32 or CKS32 is more appealing in many ways because it has a kickass ADC and a lot more I/O pins, but the STM32 can only source or sink 25 mA per pin, and can only run at 3.3 V. (I think the CKS32 is the same, but I have a harder time reading the datasheet.) And I only have two, so if I blow one up, I will be a lot sadder than if I lose a 2313 or a few ATTiny45s.
As for 45s, I have dozens, and it might be possible to gang several of them up to drive the display, maybe using two wires for I²C (“TWI” according to Atmel) and leaving four GPIOs per chip, or three GPIOs plus an ADC. Four of them would be enough to drive the display and read the keyboard, though you’d probably still want external drive transistors for the display digits.
At my house I picked up a breadboard that has two 18-pin two-digit 7-segment red displays in it, probably from 2006. They’re labeled LDD5111-11. These have decimal points and need external current limiting. They’re probably the largest digits I have available. The datasheet, from 1996, says they have a surprisingly rational pinout with 16 individual per-segment anodes and two common cathodes. You could gang up the corresponding per-segment pins to get a 12-pin interface, but you’d probably want to drive them at more than 40 mA per digit, so you probably still want external low-side drive transistors for the common cathodes.
The datasheet (“Jameco Part Number 24723”, “Ligitek”, for a whole family of such displays) says these are red GaP LEDs at 697 nm with a 90-nm bandwidth and a 1.7–2.8 V forward voltage drop, with 2.1 V being typical; 0.5 millicandela minimum at 10 mA, 0.8 millicandela typical; either 40 mA or 15 mA maximum current “per chip” (or 200 mA or 60 mA pulsed at a 10% duty cycle), and either 110 mW or 45 mW maximum power per chip; and 10 μA reverse current “absolute maximum”. I don’t know which power rating to use; the higher rating is “SR”, and the lower is “H”, but I suspect it’s the lower one, because that’s the one that is dimmer at 10 mA.
I think “per chip” means “per LED” rather than “per digit” or “per package” because 15 mA per digit would be 2 mA per LED, just too low.
This may be the easiest option of all, as well as the most readable, indoors anyway.
So, to run these manually off 5 volts, I probably want 8 per-segment resistors of 220–2200Ω, and maybe 55Ω when driving them with a microcontroller.
A little testing shows that the forward voltage at 14.5 mA is about 2.1 V, and they aren’t terribly bright at that current. I may try burning out a decimal point in one to see how bright they go.
I have a talking Kenko KK-9835TS desk clock/calculator from 2007, which has an extremely abrasion-resistant transparent keyboard, an 8-digit 7-segment LCD display, a dynamic speaker with a plastic cone, and lots and lots of wasted space inside. It runs off two AA batteries. It’s about 155×120×32 mm and weighs 143 g. The mainboard is of course a chip-on-board with a blob of epoxy over it.
This display probably has the largest LCD digits of the displays I have and would thus be optimal for daylight readability, and the flex cable to the LCD has a very comfortable 1.26-mm “pin spacing” between its 32 or so lines. The slightly mushy 26-key keyboard is also connected via a flat-flex cable, making it ideal for rebraining. The keys are easily replaced (I’ve turned the digits upside down) but it might be even more interesting to reformat them by taking them out, grinding the opaque backing off the back, and replacing it with a custom-printed sticker.
Driving 32 LCD lines in software probably requires 32 GPIOs and a certain amount of attention to properly alternating the polarity so you don’t electrochemically degrade the display. But it doesn’t require much voltage or current.
I have a Kadio KD-82TL copy of a Casio scientific calculator from 2002, which I’ve written about previously in Dercuano. It has a 50-key very mushy keyboard and a two-line display, one line of which is something like 12 5×7-pixel bitmaps (used for displaying formulas and programs) and the other line of which is something like 12 7-segment digits with decimal points. (It also has 11 indicators for calculator modes.) It’s about 85×165×30 mm and weighs 139 g with the cover on.
This is also very appealing for rebraining; again, both the keyboard and the screen are connected to the brainboard with flat-flex cables, and it runs off a couple of AA batteries, which I have removed because they had run down. It comes with a protective shield that slides over the face to protect it from damage. There is less wasted space inside than on the Kenko.
Driving the LCD seems potentially a lot more challenging than on the Kenko: there are twice as many pins, and they are less than a millimeter apart. Worse, I seem to have torn the flat-flex cable off the LCD edge where it was connected, and so I need to figure out how to re-establish contact.
This alphabetic LCD display is in a “Franklinⓡ model TES-118A Spanish-English translator, ISBN 1-56712-689-8” Beatrice and I bought in 2005. The whole device runs off a CR-2032 coin cell, measures about 105×70×15 mm, weighs 64 g, and has a pushbutton-latching domed clamshell cover that covers everything but the power button and the screen, which is recessed so far into a hole in the clamshell that it still hasn’t broken, though it’s a little scratched. Under the clamshell cover is a very mushy rubber QWERTY keyboard with arrow keys, five ideographic keys for different “applications”, and six miscellaneous buttons: menu, power, clear, back, space, and “conj”. Since I don’t have a CR-2032 handy, I’m not sure how many pixels are on the display, but it’s intended to display a few words at a time.
Opening the case was a little tricky; the clamshell unclipped from two steel rods acting as hinge pivots, three Philips screws under the battery cover came out, and then the two halves of the bottom of the case unclipped with fingernail pressure — at which point the tiny hinge pivots and plastic clamshell latch button fall out. I was glad to have a styrofoam sandwich tray underneath it to catch the parts, but still had to shake some out of my clothes.
This is unbelievably appealing for rebraining: a pocket computer ruggedized to survive being sat on, with a daylight-readable tiny LCD that can run on microwatts, with a space for a coin cell! But it would be very challenging: the LCD connector not only has many pins (about 100) but also the keyboard is on the same PCB as the chip-on-board epoxy brainblob. And there’s not a lot of wasted space; from the dimensions above you can see that it’s about 0.6 g/cc. So the most feasible approach might be to grind away the chip-on-board and wire up the new brain to the now-floating connections, rather than try to use the flat-flex connector.
There’s actually a bit of extra space under a silver-covered “keyboard template” which snaps over the keyboard. Like a millimeter or three. Hard to use but it’s about 15% of the device’s total volume.